Re: [AMBER] A General Question with regards to the NVT & NPT ensembles in MD Simulations

From: Yip Yew Mun <>
Date: Fri, 5 Jul 2013 21:16:34 +0800

Thanks David.! At least I now understand why I'm running NPT most of the time. But is there any significance in running NVT for a certain time frame, then continue with NPT, for MD simulations.? (I'm picking statistical mechanics currently. :)

Sent from my iPhone

On 5 Jul, 2013, at 9:01 PM, David A Case <> wrote:

> On Fri, Jul 05, 2013, Yip Yew Mun wrote:
>> It's kind of a newbie question, but I wish to ask what's the difference
>> between using the NVT & NPT ensembles in MD simulations, especially in
>> the case of proteins? Because I seem to be running simulations under
>> the NPT ensembles and I don't seem to know why. So I wish to know what
>> quantities am I trying to measure when I use different ensembles in MD
>> simulations?
> The problem is not that this is a "newbie" question, but rather that the
> answer is rather too general and complex for a mailing list discussion.
> For large enough systems (and a protein in water is generally "large enough")
> there is little difference between ensembles. But, as a practical matter, you
> need to use NPT simulations during equilibration of liquid systems to get the
> correct density (since the initial density won't be correct in general). Once
> equilibration is over, most properties of proteins are pretty insensitive to
> a change between NVT and NPT (or even NVE).
> A next good step would be to look at a textbook on statistical mechanics, or
> the classic text on liquid simulations by Allen & Tildesley.
> [Longer winded, somewhat more advanced:]
> (a) Arguably, NPT is closest to the physical setup of most experiments,
> but time-dependent behavior is then suspect.
> (b) The Berendsen barostat used in Amber does not given the canonical
> behavior, suppressing volume fluctuations, especially if taup is much
> greater than 1ps. (We will have better barostats in the next release).
> (c) The only sure way I know to get correct results is to run NVE
> simulations in systems large enough to approach the thermodynamic limit.
> This is rarely done.
> (d) The "correct results" referred to in part (c) assume classical
> mechanics; nuclear quantum effects being ignored are likely to be as or
> more important than are differences between ensembles.
> (e) errors in force fields probably swamp all the concerns of the
> paragraphs above.
> ...good luck...dac
> _______________________________________________
> AMBER mailing list

AMBER mailing list
Received on Fri Jul 05 2013 - 06:30:02 PDT
Custom Search